Calculated bids

x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.

In a data-driven world, the algorithm is king. From the search results you get through Google, to the product recommendations made by Amazon – even to the distribution of police officers in certain American cities – algorithms decide it all. They’re a set of instructions that computers use to solve problems: if x equals y, then divide y by z. The output of these computations might determine anything; from how many shares of a particular company to buy, to whether or not a new album is likely to reach the number one spot.

Algorithms are particularly prevalent in financial trading. Scott Patterson’s new book Dark Pools documents the creation of what he describes as “a fantastic Blade Runner trading world few outsiders could imagine, a worldwide matrix of dazzlingly complex algorithms, interlinked computer hubs the size of football fields and high-octane trading robots guided by the latest advances in artificial intelligence”.

It’s a story that’s equal parts fascinating and frightening – and it’s playing out all over again in the advertising marketplace. Programmatic buying, algorithmic trading, real-time bidding – whatever term you use – is big news in the advertising world. “Mad men are being replaced by maths men as ad agencies bid for eyeballs in fractions of a second,” wrote Simon Duke in a Sunday Times piece published in July. Later that month, as if to confirm the hype, US trade title MediaPost carried a report that Interpublic Group was looking to automate the entirety of its media buying within the next three years.

Blink and you’ll miss it
EMarketer estimates that US advertisers will spend $2bn through real-time bidding (RTB) by 2016, which works out to 28% of online banner ad budgets. In Europe – specifically the UK, Germany and France – IDC expects a similar amount will be spent on RTB by 2016: $1.8bn, or 27% of total digital display ad spend. Compare that with the figures for 2012, which were $446m and 10% respectively, and you can see that a substantial amount of growth is forecast for programmatic buying. It’s not hard to see why.

“It will eventually become the way to transact digital display advertising,” says Steve Chester, head of data and industry programmes at the Internet Advertising Bureau UK. It offers economies of scale, he says, and it chops out a lot of the admin and heavy lifting associated with person-to-person trading. That “old-school way of selling” is too slow, too cumbersome for the digital ecosystem – and, besides, there’s too much inventory to get through, says Amit Kotecha, EMEA head of marketing for Quantcast, an audience measurement and real-time advertising company. “The only way to monetise it efficiently is to sell through real-time exchanges.”

Here’s how real-time bidding works: A web user visits a web page. As the page loads up, a call goes out to all interested parties saying, “Here’s an ad impression”. Advertisers will be told what sort of impression is available, and what the profile of the user is, based on demographic, intent, interest or behavioural data. Then the bidding starts.

At Quantcast, Kotecha says: “We see 100,000 bids every second of every day.” Advertisers submit what they’re willing to pay and the highest bid wins. Except the winner only pays the second-highest price, plus a little bit on top. This is supposed to encourage bidders to be more aggressive, says Mathieu Gbetro, the group director of product and implementation for mobile network Orange. “They can bid very aggressively without taking too much risk of paying the price they’re bidding to.”

The high price of understanding
Algorithms are used to guide these trades, says Sacha Berlik, general manager Europe for DataXu, a programmatic marketing company. Explaining the algorithm’s role in the RTB process, Berlik asks us to imagine we’re the marketing director for an FMCG company. “You might be wanting to market to local dealers,” he says, “so you want to have a distribution-optimised campaign on video, mobile and display in the areas around your retailers.” The algorithm then identifies the right touchpoints the advertiser needs to use to reach their campaign target, thus delivering a “smart media buy,” says Berlik.

Indeed, smarter targeting is the main advantage for advertisers, according to the IAB’s Chester. “It’s about being able to reach the audience you want, and being able to buy by audience rather than by product. What programmatic buying allows is for sellers to more easily and better classify their audience to make it more appealing to sell, and for buyers then to come and find that audience at scale at a price that works for them.”

Whether the price works for the publisher, though, is another matter. In the book Dark Pools, Patterson notes how the rise of high-speed, automated trading reduced the spread on stocks. Similarly, Berlik says media owners are worried about the threat of “down pricing”. “There is a risk that their inventory could be commoditised,” he says.

One response to this is for media owners to set up their own private, real-time exchanges, separate from the open ones, like Orange has done. Partly, it’s about control, says Gbetro. “As long as you are private, it means you are dealing directly with someone, with a proper contract, that covers you against bad creative and stuff life that. But it’s a question of price. Whenever you build this kind of private exchange, you know the buyer is going to pay more than on the common market. They’re going to pay more for exclusivity on inventory, or to guarantee that they get to see the impression first. It’s a win-win situation for both players.”

To command a premium, though, publishers need to know more about their audience than anyone else – and that hasn’t always been the case. Advertisers, agencies and retargeters did a much better job of building up their knowledge of audiences early on by dropping cookies on publisher sites. “So, yes, at the beginning marketplaces did destroy value because publishers were not good enough at understanding what they had,” says Gbetro.

But as advertisers look to buy increasing amounts of inventory in this way, it’s a problem that publishers need to solve – and fast. For in a world of algorithmic trading, speed is everything.